You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
templatePath comes from a configmap, seems to work... can trigger test pipeline via SNS.
Upgraded spinnaker
Saw the following "SQS bootstrap" logs (here) with above config in place:
2023-12-12 18:32:49.470 INFO 1 --- [ main] c.n.s.e.p.aws.SQSSubscriberProvider : Bootstrapping SQS for SNS topic: arn:aws:sns:us-foo-1:012345678901:spinnaker2023-12-12 18:32:49.470 INFO 1 --- [ main] c.n.s.e.p.aws.SQSSubscriberProvider : Using template: /mnt/deploy.json for subscription: sqs
Ran terraform plan
Saw config drift, various queue settings adjusted
Drift appears related to spinnaker defaults for queue creation e.g. this
I updated terraform to match what Spinnaker is doing to avoid plan noise, but it raised the question of why it tried to bootstrap an existing queue.
The pubsub docs say "Spinnaker will do this part for you, provided that you specify what would be a valid name for the QueueARN in the echo-local.yml."
From reading the feature, "If the queue you've specified as a subscriber doesn't exist, Spinnaker will create it."
The docs make it sound like creating will be tried if any valid name is provided, and the code seems to say it is created as provided only if it doesn't exist. It exists, but we still saw metadata (timeouts, policy) adjusted during the upgrade which caused the drift.
Steps to Reproduce:
Unfortunately I haven't been able to repro. I've reverted the terraform changes, re-ran hal deploy, but see no spinnaker SQS bootstrap logs or terraform drift. I haven't found existing github issues with similar reports. The only suspect piece to me (and I am not a java developer so most likely missed something) is that ensureQueueExists takes more than ARNs as input (made me think it was more of a diff of queue config vs just name/exists check), but stopped digging there and it may just be expected coupling between check/create.
Additional Details:
Based on the slight disparity between docs/code and inability to find similar answer or example config in existing issues, I'd like to verify best practice when using both externally managed SQS and SNS. What should echo-local.yml look like in this case? Would we simply drop queueARN?
This issue is tagged as 'stale' and hasn't been updated in 45 days, so we are tagging it as 'to-be-closed'. It will be closed in 45 days unless updates are made. If you want to remove this label, comment:
This issue is tagged as 'stale' and hasn't been updated in 45 days, so we are tagging it as 'to-be-closed'. It will be closed in 45 days unless updates are made. If you want to remove this label, comment:
Issue Summary:
Observed SQS bootstrap log messages despite having externally-managed SQS/SNS configured in
~/.hal/default/profiles/echo-local.yml
.Cloud Provider(s):
AWS
Environment:
Feature Area:
Echo / Pubsub
Description:
Our goal is clean separation between "infra" and "service" management.
echo-local.yml
:templatePath
comes from a configmap, seems to work... can trigger test pipeline via SNS.I updated terraform to match what Spinnaker is doing to avoid plan noise, but it raised the question of why it tried to bootstrap an existing queue.
The pubsub docs say "Spinnaker will do this part for you, provided that you specify what would be a valid name for the QueueARN in the
echo-local.yml
."From reading the feature, "If the queue you've specified as a subscriber doesn't exist, Spinnaker will create it."
The docs make it sound like creating will be tried if any valid name is provided, and the code seems to say it is created as provided only if it doesn't exist. It exists, but we still saw metadata (timeouts, policy) adjusted during the upgrade which caused the drift.
Steps to Reproduce:
Unfortunately I haven't been able to repro. I've reverted the terraform changes, re-ran hal deploy, but see no spinnaker SQS bootstrap logs or terraform drift. I haven't found existing github issues with similar reports. The only suspect piece to me (and I am not a java developer so most likely missed something) is that ensureQueueExists takes more than ARNs as input (made me think it was more of a diff of queue config vs just name/exists check), but stopped digging there and it may just be expected coupling between check/create.
Additional Details:
Based on the slight disparity between docs/code and inability to find similar answer or example config in existing issues, I'd like to verify best practice when using both externally managed SQS and SNS. What should
echo-local.yml
look like in this case? Would we simply dropqueueARN
?The text was updated successfully, but these errors were encountered: